Spectral filtering is a very useful optical function that can be utilized to control the temporal waveform of pulsed optical signals, to cross-correlate or otherwise process optical signals, and to differentially control and manipulate spectrally-distinguished optical communication channels, as found for example in wavelength-division-multiplexed (WDM) optical communication systems. Devices have been introduced over the years to perform spectral filtering, all of which have characteristic shortcomings, along with their strengths. In many cases these shortcomings, including limited spectral resolution, alignment sensitivity, fabrication difficulties, high cost, and lack of flexibility, have prevented widespread application.
A spectral filtering device, according to the present usage, is a device that applies a fixed or dynamically re-programmable, complex-valued, spectral transfer function to an input signal. If Ein(ω) and Eout(ω), respectively, represent Fourier spectra of input and output signals, computed on the basis of the time-varying electric fields of the two signals, and T(ω) is a complex-valued spectral transfer function of modulus unity or smaller, the effect of the spectral filtering device can be represented as
 Eout(ω)=T(ω)·Ein(ω)
The transfer function T(ω) has an overall width Δω and a resolution width Δr, where the latter quantity is the minimum spectral interval over which T(ω) displays variation (see FIG. 1). Δr is a measure of the transformation ability of a spectral filtering device. The physical characteristics of a spectral filtering device 100 determine the range and types of spectral transfer functions that it can provide. We limit our discussion here to spectral filtering devices that act to apply a fully coherent transfer function, i. e. the device fully controls the amplitude and phase shifts applied to an input signal spectrum, except for an overall phase factor.
As a special case, if T(ω) is set equal to the conjugate Fourier spectrum Eref*(ω) of a reference temporal waveform, also called the design temporal waveform, the output field from the spectral filtering device is proportional to the cross-correlation of the input field with the reference temporal waveform. Temporal cross-correlation capability is widely useful in temporal pattern recognition.
Spectral filtering devices can be utilized to transform input signals from one format into another, or to tailor spectra to some preferred form. A spectral filtering device, according to the present usage, may or may not have the additional capacity to transform the spatial wavefront of input optical signals.
The capabilities of a spectral filtering device can be utilized in multiple ways in communications systems, including signal coding and decoding for Code-Division Multiplexing (CDM), optical packet recognition, code-based contention resolution, as WDM multiplexers and demultiplexers, and as WDM add/drop multiplexers. FIG. 2 (prior art) depicts the encoding and decoding of optical signals in a CDM context. Data 202 is input through a first communication channel, and data 206 is input through a second communication channel. Data 202 passes through a spectral filter 204, which encodes data 202 with an identifying code. Similarly, data 206 is encoded with an identifying code by a spectral filter 208. The encoded signals are combined and transmitted over an optical transmission line 210. At their destination the encoded signals are split into two paths, 212 and 214. The upper path 212 feeds into a spectral filter 216, which imparts a transfer function that is the conjugate transfer function of the filter 204. The output of spectral filter 216 is a signal comprising the superposition of data 202 and data 206; however, due to the encoding imparted by spectral filters 204 and 208 and subsequent decoding by spectral filter 216, this output signal contains a component 218 originating from 202 that has a specific recognizable temporal waveform, typically comprising a brief high power peak for each bit transmitted, along with a component 220 originating from data 206. In the upper path, the component originating from data 206 has a temporal waveform structure that can be discriminated against in detection. Typically, the component 220 originating from the data 206, has no brief high power peak.
In similar fashion, the lower branch 214 feeds into a spectral filter 222, the output of which is a signal made up of the superposition of a component 224 originating from data 206, and a component 226 originating from signal 202. As before, the two signal components have distinguishable temporal waveforms, with the component from data 206 typically having a brief detectable high power peak while the component from data 202 lacking the brief high power peak, and hence remaining below a detection threshold. An element in CDM detection is the implementation of thresholding in the detection scheme that can distinguish input pulses of differing temporal waveform character.
A variety of other CDM methods are known, many of them having need for high performance spectral filtering devices. Some alternative CDM approaches operate entirely with spectral coding. Different applications for high performance spectral filtering devices exist. Spectral filtering devices capable of accepting multiple wavelength-distinguished communication channels through a particular input port, and parsing the channels in a predetermined fashion to a set of output ports, i.e., a WDM demultiplexer, have wide application. This is especially true if the spectral filtering device is capable of handling arbitrary spectral channel spacing with flexible and controllable spectral bandpass functions.
There is another class of spectral filters wherein the entire spectral filtering function is effected through diffraction from a single diffractive structure, having diffractive elements whose diffractive amplitudes, optical spacings, or spatial phases vary along some design spatial dimension of the grating. Diffractive elements correspond, for example, to individual grooves of a diffraction grating, or individual periods of refractive index variation in a volume index grating. Diffractive amplitude refers to the amplitude of the diffracted signal produced by a particular diffraction element, and may be controlled by groove depth or width, magnitude of refractive index variation, magnitude of absorption, or other quantity, depending on the specific type of diffractive elements comprising the diffractive structure under consideration. Optical separation of diffractive elements refers to the optical path difference between diffractive elements. Spatial phase refers to the positioning as a function of optical path length of diffractive elements relative to a periodic reference waveform. The spatial variation of the diffractive elements encodes virtually all aspects of the transfer function to be applied. We refer here to diffractive structures whose diffractive elements (grooves, lines, planes, refractive-index contours, etc.) possess spatial variation representative of a specific spectral transfer function by use of the term “programmed.” Programmed diffractive structures, i.e. those structures whose diffractive elements possess spatial structure that encode a desired spectral transfer function, have only been previously disclosed in the case of surface relief grating filters, and in fiber grating filters whose diffractive elements correspond to lines (or grooves) and constant index planes, respectively. Programmed diffractive structures known in the art do not provide for the implementation of general wavefront transformations simultaneously with general spectral transformations.
Programmed surface gratings and programmed fiber gratings are encumbered with severe functional constraints. A programmed surface-grating filter has a fundamentally low efficiency when used to implement complex spectral transformations, and requires alignment sensitive free-space optical elements to function. Programmed single-mode fiber-grating filters, i.e., fiber-grating filters comprising single-mode optical fiber, produce output signals that are difficult to separate from input signals (since they can only co- or counterpropagate). Furthermore, when light is input to the device and output from the device using only the single propagating mode of the fiber, programmed single-mode fiber-grating filters can only support a single transfer function within a given spectral window.